I also forgot to mention I'm using a 256Gb SSD and have a 2TB and 1TB internal drive using SATA. I've tried exporting the file through the 1TB drive where the project/source files are on the 2TB drive and Premiere program files are on the SSD. I'm still getting over 4 hours of export time.
premiere doesn't use nvenc h264, so you wont see any help from the gtx 980 for the encode to h264. unless someone knows of some reason for the poor render time, you may have to see if something specific is causing the slow export, such as an effect on the timeline. you can use the windows resource monitor and gpu-z to help find what is causing the slow down. GPU-Z Download you could start with a duplicate of the sequence and removing all effects... also for windows 7, make sure you have service pack 1 installed.
Have you tried exporting from the timeline?
I seem to recall reading about some CC 2014 AME performance issues over in the AME forum section, but I seem to recall that exporting from the timeline may not be effected the same way.
When it is "slow", what percent of your CPU usage are you seeing in Task Manager?
Do you have any 3rd party filters, etc. on your timeline? (some can take Premiere's multi-threading capabilities and cut them back to single-threaded).
Looking at my CPU Usage in my Resource Monitor, it's only using 10-12% of the CPU during rendering and the GPU Load on GPU-Z is between 0-2%. I'm only assuming that PP is not even utilizing the GPU at all during the export. I've tried exporting both the timeline in PP and AME, but I'm still getting renders at over 4 hours for a 25 minute video with minimal visual effects on one video track(I just have an alpha glow set on it). I've tried going back and forth in the Project Settings from CUDA to Software rendering to see if it might just be a bug in PP, but still getting the same results. I'll try a different format to see if it's an issue with encoding the AVI file to h.264, but I never had this problem with CS6 and a GTX 675M card.
Download our PPBM benchmark. It has an export of a H.264 7-layer timeline that shows much GPU usage and of course lots of CPU usage. Report back here with the Output.csv results. Maybe we can see something from those results.
Okay, so I removed a simple alpha glow video effect on the video track and exported again through the timeline. I saw export time at around 4 minutes with around 40% CPU Usage on my Resource Monitor for PP and around 40% GPU Usage on GPU-Z. Does something this small really effect the total encode time? It doesn't make sense when I had the video effect on, the GPU wasn't even being used to render the entire sequence and now without the video effect, it somehow recognizes that it should be using the GPU and cuts the render time down to < 5 minutes? I'm going to try and work with other video files that I have that don't use the alpha glow and see if I get the same results.
My wild guess would be that that effect is only single threaded and with only one core working it does not feed data to the GPU.
I did a quick test in your behalf with a test project with and without "alpha glow" applied to all of the media.
The results were actually kind of strange:
- export to .mp4 utilized all cores (6-core box w/ Hyperthreads) with a smooth 82% utilization for the whole export; memory usage (also from task manager) went up and once it hit a maximum, it stayed there
- export to .mp4 with "alpha glow" applied -> CPU usage started out pretty good, then memory usage maxed out and then actually cam down; after a bit, the CPU usage stabilized at around 22%, or about 25% of what it had been at the start of the export
Smells to me like an Adobe bug or design-flaw.
Anyway, good luck with your testing.
I totally agree with Bill's suggest to do a full PPBM7 test on your new rig and report the scores back here - and to the PPBM7 site of course. This will help you to have confidence in your new hardware and setup. Once your hardware and configuration (for Win7, drive caching, Premiere settings, etc.) are working fine, then your particular workflow, etc. can be a separate troubleshooting step.